UDK

source code of Universally Describing Keypoints From a Semi-Global to Local Perspective, Without Any Specific Training

Keypoint extraction represents fundamental and pivotal tasks within the realm of robotic vision. Deep learning-based approaches for keypoint extraction showcase remarkable prowess and achievements. Nevertheless, these deep learning methodologies often necessitate specialized training on extensive datasets. While methods based on pre-trained backbone network feature maps for feature matching exist, they are still constrained by specific algorithmic workflows that do not decouple keypoints' descriptors from the matching process. We study the method of relying exclusively on a pre-trained backbone network sourced from ImageNet for keypoint extraction and demonstrate the impact of different detection strategies and descriptor composition strategies on matching performance. The proposed pipeline obviates the need for tailored training while concurrently achieving state-of-the-art performance. To validate the efficacy of our algorithm, comprehensive evaluations are conducted across the HPatches, MegaDepth, and Scannet datasets.

requirments

pytorch >= 1.8.0
numpy
opencv-python >= 4.4.0
argparse
PyYaml
matplotlib
loguru
glob
h5py
kornia
pathlib
tqdm

Test Data

HPatches

Please remove the following folders after downloading: i_contruction i_crownnight i_dc i_pencils i_whitebuilding v_artisans v_astronautis v_talent.

megadepth-1500 and scannet-1500

Evaluation

python dfm.py

The settings of various algorithms are recorded in configs/xxx.yml, dfm.py, and ManyDeepFeatureMatcher.py.

By the way, the ManyDeepFeatureMatcher.py mainly controls the detection strategies, detection params, and the level of multi-scale.

We will optimize the parameters structure as soon as possible.

We would like to thank the SuperGlue, LoFTR and DFM authors and contributors for making their codes or data open source which inspired.